Switch the model for a tab

Agents: claude, codex, opencode, cursor, devin, copilot, muse · On: phone, desktop

Each tab runs one model. Change it by tapping the model button under the composer, or by typing /model.

Why you would use it

The tab was opened for a quick question on the cheap model and has turned into real work. Or the reverse: you are on the expensive model and about to ask something trivial. The model is per tab, so one tab can grind on a big refactor while another answers quick questions.

How to use it

Tap the model button (#modelpick) in the row under the composer, between the agent's name and the effort. It shows the model this tab runs, or auto.
The slash panel opens on its Models slot: a search bar reading Search models and Claude · 5, and the list of models under the agent's name, with a green tick on the one this tab is already using. Tap another, or type in the search bar to filter. A search also lists the matching models of every other agent on this box, each under its agent's name, and the count reads All. Tapping one hands the tab to that agent on that model, at auto effort. On a model with effort levels, the Effort slider sits over the search bar; see set the effort. On a model with fast mode, the Fast switch sits under it; see fast mode.
Choose the model and effort, then star the model to save its harness, model and effort in Favourites. Your favourites lead the Models list under Favourites, from every harness, so one tap restores a saved configuration. The Favourites slot lists them too. Each favourite shows its effort on the right: the level's name and five rising bars, grey for low up through green, yellow and orange to red for max. auto leaves the bars unlit. A star on another model saves it at auto effort. The house beside the star makes a model the one new tabs open on; see default-start. See the slash menu.

The list is what the agent's own CLI offers on this machine, so a new model shows up without an update to VCode. Type /refresh-models to ask the CLIs again now; see update-from-a-tab.

Or, by typing:

Type /model. The menu becomes the same list of models as soon as the word is finished (no space needed) and keeps filtering as you type a value, across every agent the same way. Typing /model while the panel is already open from the model button hands the panel to the draft. It works at the end of a message already written: picking a model there sets it and takes /model out of the draft, leaving the message.
Send /model on its own to see what the tab is on instead.
Type /model sonnet and send it to set it directly. The name the menu shows works too: /model Sonnet resolves to sonnet.

Both the picker and the typed command go through the same place, so they cannot mean different things.

What you see

The button's label is the model's product name, not a caption: Opus 5.5, GPT-5.6 Sol, or auto in a dimmed style when nothing is set.
Long ids are shortened for the label: claude-opus-5-5[1m] reads as Opus 5.5, a trailing snapshot date is dropped, and gpt-5.6-sol reads as GPT-5.6 Sol.
Each row shows the model's name and the note its CLI gives it: Sonnet, Sonnet 5 · Efficient for routine tasks.
The model this tab is on carries a green tick on the right of its row. It follows what the tab is really running: auto holds the tick only until the first turn, after which the agent reports the id it started with and that model's row takes it.
A full model id is ticked on its row. Running claude-fable-5-1 ticks Fable.
Only one row is ticked. When the menu lists both Opus 5.5 and Opus 5, running claude-opus-5 ticks Opus 5 alone. A release with no row of its own ticks its family's alias row.
The tooltip is Model: opus, or Model: whatever the CLI is configured for when nothing is set.
/model with no argument answers with a toast: model: opus, or model: default.
A second tap on the button closes the menu.

Options and settings

Every agent's list comes from its CLI on this machine, as the account it is logged into sees it. Each list ends with auto (whatever the agent is configured for), which sends no model and leaves the choice to the CLI.

Agent	Where the list comes from
claude	The models the CLI offers this login, with the names and notes it gives them (`opus[1m]` is `Opus (1M context)`)
codex	The models the CLI lists for this account; hidden ones are left out
opencode	The models of the providers it is logged into
Cursor	The model IDs and names ACP advertises; the picker sends the exact ID, and `auto` uses ACP's `default[]`
devin	The models `devin models list` names, narrowed to the ones your account's session allows. Each effort is a variant of the model, and the picker sends the variant's ID
copilot	The models in the model section of `copilot help config`, with Auto first. Every model but Auto takes the efforts `none`, `minimal`, `low`, `medium`, `high`, `xhigh` and `max`
muse	The models `model/list` names on a `muse serve` of its own, with each model's efforts as its variants. Logged out, Muse lists none

The box asks each CLI once and keeps the answer for an hour. After that the next look at the list is served the old answer while the box asks again in the background. /refresh-models asks at once. Devin's list is kept until /refresh-models asks again; the list of Devin's families is kept for 10 minutes.

Limits and known gaps

Tapping the ticked row is not a no-op: it sends the row's id again, and on claude that restarts the agent process.
A CLI that is missing, logged out or slow to answer gives an empty list. The menu then offers auto alone, and the phone asks again a few times.
The setting is per tab, not global. A new tab opened with /new copies this tab's model.
The value sent to the CLI is the row's id (opus[1m]), not its name; the row's tooltip shows what will be sent.
On claude and opencode the model is a process flag, so the agent process is stopped and the next turn starts a new one. Nothing is lost: that start resumes the same vendor session.
A change made mid-turn is held. The process that owns the turn keeps it, and the restart happens before the next turn.
On devin a model change restarts the agent process, and the next load sets the model again.
A model your Devin plan cannot run is refused with devin cannot run <model id> on this account; it offers <model ids>. The Free plan runs one model.
On copilot a model or effort change restarts the agent process. Copilot takes both when it starts, so a change starts it again on the same session.
On copilot a model your plan lacks is replaced. Copilot runs another model in its place without saying so.
On muse a model or effort change applies to the next message, with no restart.
On codex the model rides on turn/start as an override, so it lands on the very next message with no restart.
A new model reaches the menu within the hour after its CLI offers it. An update to the CLI itself comes from /update, or from the CLI's own updater.
The button is disabled with no tab open.
Choosing a model from the menu does not clear your half-written message. Only when the draft is the command (/model …) is the box emptied.